The accompanying slides are here
In this demo will be using OpenAi Gym, a standard API for reinforcement learning with a lot of built in environments
- Lets create a new virtual enviorment to house our new project called OpenAiGym by typing the following comand into the terminal
conda create -n uais-rl python=3.7 - Next we will active our enviorment
conda activate uais-rl - If you do not have miniconda installed you can get it here
- Lets create a new virtual enviorment to house our new project called OpenAiGym by typing the following comand into the terminal
python3 -m venv OpenAiGym-env - To activate on Windows run:
OpenAiGym-env\Scripts\activate.bat - To activate on Unix or MacOS run:
source OpenAiGym-env/bin/activate
- Firstly lets install Jupyter Notebook
pip install notebook - First we need to install the following package
pip install --user ipykernel - Next we need to add the kernel so we can have it in our Jupyter Notebook
python -m ipykernel install --user --name=uais-rl - Later if you wanna remove the enverment use
jupyter kernelspec uninstall myenv
- Next we need to install the base gym library
pip install gym - We will also need to install the atari enviorment dependences
pip install 'gym[atari]' - You can freely download Atari 2600 roms here but the Breakout ROM that we will be using is provided
- Next we will use ALE to import our ROM
ale-import-roms ROMS/ - Next install imageio for capturing our image frames
pip install imageio - and lastly install cv2
pip install opencv-python
- Next install pytorch
conda install pytorch -c pytorch - Next clone this repo
- and then install all the requirements
pip install -r requirements.txt - lastly
pip install 'stable-baselines3[extra]'